Ozone Depletion in Tropospheric Volcanic Plumes: From Halogen-Poor to Halogen-Rich Emissions

: Volcanic halogen emissions to the troposphere undergo a rapid plume chemistry that destroys ozone. Quantifying the impact of volcanic halogens on tropospheric ozone is challenging, only a few observations exist. This study presents measurements of ozone in volcanic plumes from K¯ılauea (HI, USA), a low halogen emitter. The results are combined with published data from high halogen emitters (Mt Etna, Italy; Mt Redoubt, AK, USA) to identify controls on plume processes. Ozone was measured during periods of relatively sustained K¯ılauea plume exposure, using an Aeroqual instrument deployed alongside Multi-Gas SO 2 and H 2 S sensors. Interferences were accounted for in data post-processing. The volcanic H 2 S/SO 2 molar ratio was quantiﬁed as 0.03. At Halema‘uma‘u crater-rim, ozone was close to ambient in the emission plume (at 10 ppmv SO 2 ). Measurements in grounding plume (at 5 ppmv SO 2 ) about 10 km downwind of Pu‘u ‘ ¯O‘¯o showed just slight ozone depletion. These K¯ılauea observations contrast with substantial ozone depletion reported at Mt Etna and Mt Redoubt. Analysis of the combined data from these three volcanoes identiﬁes the emitted Br/S as a strong but non-linear control on the rate of ozone depletion. Model simulations of the volcanic plume chemistry highlight that the proportion of HBr converted into reactive bromine is a key control on the efﬁciency of ozone depletion. This underlines the importance of chemistry in the very near-source plume on the fate and atmospheric impacts of volcanic emissions to the troposphere. were performed to determine H 2 S/SO 2 and ∆ O 3 / ∆ SO 2 using robust-ﬁt algorithms that yield similar results to least-squares but are less affected by outliers. The reported linear model trends exhibit p -values < 0.05 that indicate signiﬁcant explanatory power. The R 2 coefﬁcient of determination (ratio of the explained variation to the total variation) was calculated using coefﬁcients of correlation and is also provided in the Results.


Introduction
Volcanoes release large quantities of gases and aerosols to the atmosphere. Very large explosive eruptions inject gases directly to the stratosphere, but a significant number of smaller eruptions and continuously passive degassing volcanoes release their emissions to the troposphere; the total global SO 2 flux from passive degassing over 2004-2016 was recently estimated as 23 Tg/yr on average, exceeding volcanic eruptive emissions by about one order of magnitude [1]. To date, studies of the atmospheric chemistry and climate impacts of volcanic emissions have mostly focused on SO 2 and its oxidation to sulfate particles in both the stratosphere (e.g., [2]) and troposphere (e.g., [3]). Volcanic sulfates also catalyse gas-aerosol reactions leading to reductions in stratospheric ozone levels, e.g., [4,5]. However, the volcanic release contains a number of other gases and particles, including notably the emission of volcanic halogens such as HBr and HCl (e.g., [6]). These were initially assumed to be simply washed out of the plume in the troposphere and deposited. Volcanic halogens can occasionally be injected into the stratosphere as evidenced by recent observations of HCl, OClO, BrO, and IO by satellite [7][8][9][10] and of HCl and HF by an instrumented aircraft that transected a high-altitude volcanic cloud [11]. Volcanic halogens that reach high altitudes may cause reductions in stratospheric ozone levels. This has been both observed [11] and simulated by numerical models (e.g., [12,13]) using atmospheric chemistry schemes that were originally developed to study impacts from anthropogenic sources of halogens (chlorofluorocarbons, CFCs) on stratospheric ozone.
The goals of this study are: (i) to quantify the H 2 S emission alongside SO 2 (i.e., H 2 S/SO 2 ratio) enabling the prediction of cross-sensitivity on the Aeroqual ozone measurement; (ii) to measure ozone in the crater-rim plume; (iii) to measure ozone in the chemically more evolved downwind plume; and (iv) to interpret the ozone observations in the context of volcanic plume halogen chemistry and tropospheric ozone depletion reported in volcanic plumes globally. Section 2 below outlines the challenges to measuring ozone in volcanic plumes; Section 3 describes the instruments and field-deployment; Section 4 presents the ozone measurements made at Kīlauea; and Section 5 discusses these observations in a wider volcanic plume chemistry context.

Interferences of Volcanic Gases on the Ozone Measurement
Ozone is present in the background troposphere at mixing ratios of tens of nmol/mol (or ppbv). Two main in-situ approaches to measuring atmospheric ozone are ultra-violet (UV) spectroscopy on ground-based or aircraft platforms (with instruments often operating at 254 nm, within the Hartley ozone absorption band) and the ozone (electrochemical) balloon sonde. A challenge to measuring ozone in volcanic plumes is the presence of other interfering gases at much higher abundances (e.g., µmol/mol or ppmv) than typically occur in the background atmosphere. These can induce positive or negative interferences to yield an erroneously high or low ozone measurement, respectively.
Specifically, volcanic SO 2 induces a negative interference on electrochemical cell measurements of ozone. For example, ozone-sondes launched into the Eyjafjallajöjkull 2010 eruption plume over Europe detected severely disturbed profiles [31], but the low ozone signal could not be quantifiably attributed to volcanic plume chemistry due to the interference from SO 2 . Volcanic SO 2 induces a positive interference on UV spectroscopic measurements of ozone due to its absorption at 254 nm. This interference can be automatically corrected in dual channel instruments that contain a second (ozone scrubbed) channel, or may be subtracted in data post-processing using co-measured SO 2 , provided that the plume is sufficiently dilute [23]. Several airborne measurements of ozone in volcanic plumes have been reported; ozone concentrations were observed below ambient levels in a study of volcanic plumes predominantly from Alaska [32]. Ozone was depleted by at least a third compared to background, reaching up to 90% depletion in the 1980 Mount St. Helens eruption plume, during both ash-rich and ash-poor (passive degassing following the eruption) conditions [33,34]. Oppenheimer et al. [35] reports ozone depletion up to 35% with respect to background in the Mt Erebus (Antarctica) plume, alongside measurements of volcanic sulfur and nitrogen species, but also with some evidence for a very rapid (and unexplained) near-source reduction in SO 2 . Instrumented aircraft campaigns by Vance et al. [36] and Schumann et al. [37] measured ozone depletion alongside volcanic SO 2 and other gases and particles in the 2010 Eyjafjallajokull eruption plume dispersed over Europe. Volcanic BrO was also observed in the downwind plume. However, the total volcanic bromine emission from Eyjafjallajokull was not well monitored. Finally, an instrumented aircraft campaign spatially mapped ozone depletion alongside SO 2 in the 2010 Mt Redoubt eruption plume over 2-20 km downwind [27]. In this study, Kelly et al. also quantified the volcanic bromine emission by filterpack (HBr/SO 2 = 4.1 × 10 −3 mol/mol), enabling interpretation of the observations and a comparison to an atmospheric box model. Their study captured a rapid decrease in ozone in the near-downwind plume reaching up to tens of ppbv O 3 loss, followed by a slow (partial) recovery in the dispersing plume. This demonstrated the need to characterise plume ozone-depleting chemistry over short spatial-timescales very near to the source. To do so, diffusion tubes were installed on Mt Etna flanks by Vance et al. [36], who reported an ozone depletion signature that was anti-correlated with SO 2 . However, corrosion problems prevented a measurement very close to the crater-rim. Some initial measurements were also made using a UV ozone instrument with CrO 3 scrubber [36]. This approach was significantly furthered by Surl et al. [24], whose novel observations quantified the rate of ozone depletion through measurements on Mt Etna's flanks. Their approach involved a portable single-channel UV spectroscopic instrument combined in series with CrO 3 scrubbers to remove SO 2 from the air inlet, with the ozone losses on the scrubber quantified and treated in the data post-processing. This enabled-for the first time-quantification of ozone depletion in very near-to-source plumes containing tens of ppmv of volcanic SO 2 that was co-measured by small electrochemical sensor (Multi-Gas instrument). The volcanic bromine emission was also characterised during the field-campaign (HBr/SO 2 = 6.13 × 10 −4 mol/mol). Through observations made at up to several hundred meters distance from the summit emission source (equivalent to up to a few minutes travel time), Surl et al. [24] derived a linear equation for ozone loss in Mt Etna's near-source plume as a function of travel time downwind from the summit craters, finding a gradient of ∆O 3 /∆SO 2 = (−1.02 ± 0.07) × 10 −5 s −1 and intercept of (−6.2 ± 0.05) × 10 −4 mol/mol. The non-zero intercept suggests that some depletion of ozone had already occurred before the gases reached the crater-rim. The ozone depletion rate (gradient) has been compared to atmospheric 1D and box model simulations by Surl et al. [24] and Roberts et al. [30].
Here, we used an Aeroqual instrument (WO 3 sensor; Aeroqual Limited, Auckland, New Zealand) to measure ozone in the Kīlauea volcanic plumes (Halema'uma'u and Pu'u 'Ō'ō, HI, USA). The instrument is much less sensitive to interference from volcanic SO 2 than other methods, see Table 1. However, the sensor exhibits other cross-sensitivities as listed in Table 1. The most important cross-sensitivity for volcanic plumes is that of H 2 S at −2.5% that needs to be considered in data post-processing. The importance of this interference depends on the magmatic conditions specific to each volcano. Fumarolic emissions are H 2 S-rich (e.g., Vulcano, with H 2 S/SO 2 molar ratios of 1 [38]). Oxidized magma emissions are H 2 S-poor and dominated by SO 2 (e.g., Masaya, Mt Etna, Kīlauea with H 2 S/SO 2 molar ratios of just a few percent [39]). The H 2 S-poor nature of the Kīlauea volcanic emission is further verified by our in-situ real-time sensing of H 2 S and SO 2 alongside ozone (see Results). Alongside our field-measurements of ozone, the bromine content of Kīlauea emissions was characterized by Mather et al. [40], finding HBr/SO 2 = 2.9 × 10 −5 mol/mol.

Intermittent Plume Exposure and Sensor Response Times
In-situ measurements of SO 2 at the volcano crater-rim typically exhibit a high temporal variability. Rapid variations in local gas concentrations are often observed near to the volcanic source due to the complex local wind fields that advect the volcanic gases towards and away from the sensors. To characterize the emissions, the correlation in the time-series from two sensors (e.g., SO 2 and H 2 S) is analyzed to derive a gas ratio, e.g., H 2 S/SO 2 . However, the response times of Multi-Gas and other sensors are not all identical, which can introduce uncertainty in the volcanic gas measurement [42]. Rapid variations in gas concentrations are also a challenge to the Aeroqual instrument that makes measurements of ozone at one-minute resolution. As mentioned in Section 2.1 above, the instrument sensor is sensitive to other gases, including (volcanic) H 2 S. These cross-sensitivities have been individually quantified by Aeroqual in laboratory experiments at constant gas abundance, Table 1. However, field-deployment of the Aeroqual instrument at volcanoes can expose it to time-varying abundances of volcanic gases. Very large and rapid temporal variations in interfering gas abundances may lead to anomalous values in the ozone measurement that are difficult to correct in data post-processing (see Section 3, Materials and Methods). A potential solution is to attempt to perform measurements under conditions of more sustained exposure to the volcanic plume. Such conditions are rather uncommon due to the local wind-field conditions near-to-source and because buoyant volcanic plumes tend to become elevated above ground. Here, we present in-situ observations of SO 2 , H 2 S, and ozone in the near-source and near-downwind Kīlauea (Halema'uma'u and Pu'u 'Ō'ō, HI, USA) plumes obtained during two rare periods of relatively sustained plume exposure (>10 min at "plume strengths" of 5-10 ppmv SO 2 ) on 3 September 2007 and 19 July 2008.

Materials and Methods
The Aeroqual portable ozone monitor is a hand-held instrument for measuring ozone at low mixing ratios at one-minute resolution using a tungsten oxide (WO 3 ) sensor. The range is 0 to 150 ppbv and a reported resolution of 1 ppbv, accuracy <±5 ppbv (datasheets: www.Aeroqual. com [41]). The WO 3 sensor is highly sensitive to ozone, but a major challenge to its use for atmospheric ozone measurements is sensitivity drift. The Aeroqual instrument uses propriety software to correct for WO 3 sensitivity drift that involves different phases of sensor operation during the one-minute measurement period. Cross-sensitivities to the ozone measurement are reported by Aeroqual for fixed gas concentrations (such as for H 2 S, Table 1), however, the interferences caused by time-varying gas exposure under time-varying sensor operation are more complex to characterize or quantitatively account for in data post-processing. Therefore, this study focuses on Aeroqual measurements during some (rare) periods of relatively sustained plume exposure, and relies on co-measurements of SO 2 and H 2 S to aid interpretation of the Aeroqual ozone observations. Volcanic gases were measured at 1 Hz resolution using the pumped Multi-Gas system described by Roberts et al. [43] that includes Alphasense Ltd (Essex, UK) electrochemical sensors SO 2 -AF and H 2 S-A1 to measure SO 2 and H 2 S respectively. Reported resolution of the sensors is <0.1 ppmv (SO 2 ), <0.05 ppmv (H 2 S), and response time is <35 s (www.Alphasense.com [44]). The sensor current output was converted into ppmv mixing ratio time-series using sensitivities and cross-sensitivities determined from laboratory calibrations prior to the fieldwork. The H 2 S-A1 sensor exhibits a positive interference to SO 2 that was subtracted in the data post-processing (using SO 2 simultaneously measured by the SO 2 -AF sensor and calibration (cross)-sensitivities). Further details on the instrument and data analysis are provided by Roberts et al. [43].
Kīlauea volcano was emitting three plumes during the fieldwork: an emission from the Halema'uma'u summit crater (measured prior to the 2008 appearance of a lava-lake), the emission from Pu'u 'Ō'ō (east rift zone vents), and an emission resulting from lava contact with sea-water (not sampled). Field-measurements were made at relatively sustained plume exposure on 3 September 2007 to sample Halema'uma'u crater-rim emissions and on 19 July 2008 sampling grounding plume downwind from Pu'u 'Ō'ō, on chain of craters road. This latter site is about 10 km distance from the Pu'u 'Ō'ō vent emission source. Linear regressions were performed to determine H 2 S/SO 2 and ∆O 3 /∆SO 2 using robust-fit algorithms that yield similar results to least-squares but are less affected by outliers. The reported linear model trends exhibit p-values < 0.05 that indicate significant explanatory power. The R 2 coefficient of determination (ratio of the explained variation to the total variation) was calculated using coefficients of correlation and is also provided in the Results.

Volcanic SO 2 , H 2 S, and Ozone Measured in the Halema'uma'u Crater Rim Emission Plume
Time-series of volcanic SO 2 and H 2 S gas abundances measured at Halema'uma'u crater-rim are well-correlated (p-value < 0.05 and R-squared = 0.52), and linear-regression finds a H 2 S/SO 2 molar ratio of 0.030, see Figure 1. This result confirms the low H 2 S content of the Kīlauea plume, although the H 2 S/SO 2 molar ratio is somewhat higher than has been reported previously [39]. This might reflect changing magmatic conditions over time. The scatter in the data is most likely due to instrument response times [41]. Nevertheless, a H 2 S measurement is still possible, due to the relatively sustained exposure (e.g., around 10 ppmv SO 2 for tens of minutes). The statistical uncertainty in the gas ratio is very low (<10 −3 ) due to the very large number of data points. The actual measurement uncertainty is probably somewhat higher as discussed in Section 4.2.

Volcanic SO2, H2S, and Ozone Measured in the Halema'uma'u Crater Rim Emission Plume
Time-series of volcanic SO2 and H2S gas abundances measured at Halema'uma'u crater-rim are well-correlated (p-value <0.05 and R-squared = 0.52), and linear-regression finds a H2S/SO2 molar ratio of 0.030, see Figure 1. This result confirms the low H2S content of the Kīlauea plume, although the H2S/SO2 molar ratio is somewhat higher than has been reported previously [39]. This might reflect changing magmatic conditions over time. The scatter in the data is most likely due to instrument response times [41]. Nevertheless, a H2S measurement is still possible, due to the relatively sustained exposure (e.g., around 10 ppmv SO2 for tens of minutes). The statistical uncertainty in the gas ratio is very low (<10 −3 ) due to the very large number of data points. The actual measurement uncertainty is probably somewhat higher as discussed in Section 4.2. The volcanic SO2 time-series was filtered to make one-minute averaged data and these are compared to the ozone time-series (1 min resolution) in Figure 2. The measured ozone abundance fluctuates around 20 to 40 ppbv with a few (anomalous) zero data points. No clear trend is visible in the time-series although a scatter plot of the SO2, and ozone data show a weak anti-correlation (pvalue <0.05 and R-squared = 0.13). The gradient of the linear regression is -0.72 (± 0.13) × 10 −3 mol/mol. This measured ΔO3/ΔSO2 closely matches the predicted Aeroqual instrument response to a 2.5% H2S interference for volcanic plume with H2S/SO2 = 0.03 mol/mol (−2.5/100 × 0.03 = −0.75 × 10 −3 mol/mol). Accounting for the H2S interference leads to the deduction in that actual in-plume ozone concentrations were very close to ambient levels in the emission plume at Halema'uma'u crater rim. The volcanic SO 2 time-series was filtered to make one-minute averaged data and these are compared to the ozone time-series (1 min resolution) in Figure 2. The measured ozone abundance fluctuates around 20 to 40 ppbv with a few (anomalous) zero data points. No clear trend is visible in the time-series although a scatter plot of the SO 2 , and ozone data show a weak anti-correlation (p-value < 0.05 and R-squared = 0.13). The gradient of the linear regression is −0.72 (± 0.13) × 10 −3 mol/mol. This measured ∆O 3 /∆SO 2 closely matches the predicted Aeroqual instrument response to a 2.5% H 2 S interference for volcanic plume with H 2 S/SO 2 = 0.03 mol/mol (−2.5/100 × 0.03 = −0.75 × 10 −3 mol/mol). Accounting for the H 2 S interference leads to the deduction in that actual in-plume ozone concentrations were very close to ambient levels in the emission plume at Halema'uma'u crater rim.

Volcanic SO2, H2S, and Ozone Measured in the Near-Downwind Plume from Pu'u 'Ō'ō
Time-series of volcanic SO2 and H2S gas abundances measured in the plume ~10 km downwind from Pu'u 'Ō'ō are also well-correlated (p-value <0.05 and R-squared = 0.50), and linear-regression finds a H2S/SO2 molar ratio of 0.034, Figure 3. Visually, the data appear less scattered compared to those in Figure 1, reflecting improved instrument performance when exposed to more slowly fluctuating gas concentrations. However, instrument noise becomes important at these low H2S abundances in relatively dilute plume, leading to similar R 2 as for Figure 1. The gas ratio is slightly higher than found for the crater-rim emissions from the Halema'uma'u (H2S/SO2 = 0.030 mol/mol) which could be due to (i) slight differences in the H2S/SO2 emitted from Pu'u 'Ō'ō and Halema'uma'u despite their common magma source; (ii) partial atmospheric oxidation of SO2 in the Pu'u 'Ō'ō plume by up to 14% (this would indicate a relatively fast in-plume SO2 oxidation); or (iii) uncertainty in the Multi-Gas measurement.

Volcanic SO 2 , H 2 S, and Ozone Measured in the Near-Downwind Plume from Pu'u 'Ō'ō
Time-series of volcanic SO 2 and H 2 S gas abundances measured in the plume~10 km downwind from Pu'u 'Ō'ō are also well-correlated (p-value < 0.05 and R-squared = 0.50), and linear-regression finds a H 2 S/SO 2 molar ratio of 0.034, Figure 3. Visually, the data appear less scattered compared to those in Figure 1, reflecting improved instrument performance when exposed to more slowly fluctuating gas concentrations. However, instrument noise becomes important at these low H 2 S abundances in relatively dilute plume, leading to similar R 2 as for Figure 1. The gas ratio is slightly higher than found for the crater-rim emissions from the Halema'uma'u (H 2 S/SO 2 = 0.030 mol/mol) which could be due to (i) slight differences in the H 2 S/SO 2 emitted from Pu'u 'Ō'ō and Halema'uma'u despite their common magma source; (ii) partial atmospheric oxidation of SO 2 in the Pu'u 'Ō'ō plume by up to 14% (this would indicate a relatively fast in-plume SO 2 oxidation); or (iii) uncertainty in the Multi-Gas measurement.

Volcanic SO2, H2S, and Ozone Measured in the Near-Downwind Plume from Pu'u 'Ō'ō
Time-series of volcanic SO2 and H2S gas abundances measured in the plume ~10 km downwind from Pu'u 'Ō'ō are also well-correlated (p-value <0.05 and R-squared = 0.50), and linear-regression finds a H2S/SO2 molar ratio of 0.034, Figure 3. Visually, the data appear less scattered compared to those in Figure 1, reflecting improved instrument performance when exposed to more slowly fluctuating gas concentrations. However, instrument noise becomes important at these low H2S abundances in relatively dilute plume, leading to similar R 2 as for Figure 1. The gas ratio is slightly higher than found for the crater-rim emissions from the Halema'uma'u (H2S/SO2 = 0.030 mol/mol) which could be due to (i) slight differences in the H2S/SO2 emitted from Pu'u 'Ō'ō and Halema'uma'u despite their common magma source; (ii) partial atmospheric oxidation of SO2 in the Pu'u 'Ō'ō plume by up to 14% (this would indicate a relatively fast in-plume SO2 oxidation); or (iii) uncertainty in the Multi-Gas measurement.  The one-minute averaged SO 2 time-series is compared to the ozone measurement (at 1 min resolution) in Figure 4. The measured ozone abundance fluctuates between 30 and 50 ppbv. The last 20 min of the time-series appear mostly plume-free (low SO 2 ) so show the natural variability in ozone abundance. There were no (anomalous) zero data points. A clear tendency (seen particularly in the first 50 min) is that ozone decreases in conjunction with maximum peaks in SO 2 . A scatter plot of the SO 2 and ozone data confirms this anti-correlation (p-value < 0.05 and R-squared = 0.30). The R 2 is higher than found at the crater-rim (Figure 2), i.e., the anti-correlation of ozone to SO 2 explains a greater ratio of the variation in ozone compared to total variation for the Pu'u 'Ō'ō downwind plume measurements than for that of the Halema'uma'u crater-rim measurements. The gradient of the linear regression is −1.15 (±0.18) × 10 −3 mol/mol. This measured ∆O 3 /∆SO 2 exceeds the Aeroqual instrument response to volcanic H 2 S calculated as −0.84 × 10 −3 mol/mol for a plume with H 2 S/SO 2 molar ratio of 0.034. Accounting for the negative H 2 S interference on measured ozone at H 2 S/SO 2 = 0.034 suggests an actual in-plume ∆O 3 /∆SO 2 of −0.31 (±0.18) × 10 −3 mol/mol. 20 minutes of the time-series appear mostly plume-free (low SO2) so show the natural variability in ozone abundance. There were no (anomalous) zero data points. A clear tendency (seen particularly in the first 50 minutes) is that ozone decreases in conjunction with maximum peaks in SO2. A scatter plot of the SO2 and ozone data confirms this anti-correlation (p-value <0.05 and R-squared = 0.30). The R 2 is higher than found at the crater-rim (Figure 2), i.e., the anti-correlation of ozone to SO2 explains a greater ratio of the variation in ozone compared to total variation for the Pu'u 'Ō'ō downwind plume measurements than for that of the Halema'uma'u crater-rim measurements. The gradient of the linear regression is -1.15 (± 0.18) × 10 −3 mol/mol. This measured ΔO3/ΔSO2 exceeds the Aeroqual instrument response to volcanic H2S calculated as −0.84 × 10 −3 mol/mol for a plume with H2S/SO2 molar ratio of 0.034. Accounting for the negative H2S interference on measured ozone at H2S/SO2 = 0.034 suggests an actual in-plume ΔO3/ΔSO2 of −0.31 (± 0.18) × 10 −3 mol/mol. A source of uncertainty in the ozone depletion measurement is variation in background ozone, which appears to descend then ascend during the time-period, and also exhibits short-term variability. Additional uncertainties can arise from uncertainty in the interference from H2S, for example, an error of ±0.005 in the measured H2S/SO2 molar ratio would induce an uncertainty in the detected ΔO3/ΔSO2 of ±0.125 × 10 −3 mol/mol, whilst an error of ±0.01 would induce an uncertainty in ΔO3/ΔSO2 of ±0.25 × 10 −3 mol/mol. Nevertheless, results from the crater-rim observations suggested that H2S/SO2 measured by the Multi-Gas sensors is fully consistent with the expected interference. In summary, the data show that ozone in the downwind Pu'u 'Ō'ō plume was likely slightly depleted below ambient levels but only by a small magnitude.

Discussion and Conclusion: Ozone Depletion from Halogen-Poor to Halogen-Rich Volcanic Plumes
Measurements of ozone in volcanic plumes are challenging. This study demonstrates the need to consider the interference of H2S on Aeroqual measurements of ozone, the need for sustained plume exposure (as 1 min measurements can be affected by rapid changes in gas concentration), and also illustrates natural variability in background ozone that can act to mask an observable ozone depletion signature. Nevertheless, the Aeroqual measurements provide a useful constraint on the magnitude of ozone depletion in the Kīlauea volcanic plumes at short distances from the source. In particular, A source of uncertainty in the ozone depletion measurement is variation in background ozone, which appears to descend then ascend during the time-period, and also exhibits short-term variability. Additional uncertainties can arise from uncertainty in the interference from H 2 S, for example, an error of ±0.005 in the measured H 2 S/SO 2 molar ratio would induce an uncertainty in the detected ∆O 3 /∆SO 2 of ±0.125 × 10 −3 mol/mol, whilst an error of ±0.01 would induce an uncertainty in ∆O 3 /∆SO 2 of ±0.25 × 10 −3 mol/mol. Nevertheless, results from the crater-rim observations suggested that H 2 S/SO 2 measured by the Multi-Gas sensors is fully consistent with the expected interference. In summary, the data show that ozone in the downwind Pu'u 'Ō'ō plume was likely slightly depleted below ambient levels but only by a small magnitude.

Discussion and Conclusion: Ozone Depletion from Halogen-Poor to Halogen-Rich Volcanic Plumes
Measurements of ozone in volcanic plumes are challenging. This study demonstrates the need to consider the interference of H 2 S on Aeroqual measurements of ozone, the need for sustained plume exposure (as 1 min measurements can be affected by rapid changes in gas concentration), and also illustrates natural variability in background ozone that can act to mask an observable ozone depletion signature. Nevertheless, the Aeroqual measurements provide a useful constraint on the magnitude of ozone depletion in the Kīlauea volcanic plumes at short distances from the source. In particular, the observations provide a valuable end-member example that quantifies ozone in the plume of a low halogen emitter. This contrasts to studies to date that mostly focused on volcanic plumes with higher halogen contents. To interpret the field-data, ozone depletion is scaled relative to co-measured SO 2 in order to distinguish between chemistry and plume dilution effects, and it is interpreted in terms of distance or travel time downwind and the bromine emission.
For the Kīlauea plumes, ozone was shown to be close to ambient levels during plume exposures of up to 10 ppmv SO 2 at the Halema'uma'u crater-rim, and they were only slightly depleted in the grounding plume at around 10 km downwind from Pu'u 'Ō'ō (e.g., by up to 2 ppbv measured at 5-8 ppmv SO 2 ). The measurements indicate that the volcano emission has only a limited impact on tropospheric ozone, as might be expected for a low halogen emitter. This contrasts with findings for higher halogen emitters. A study of Mt Etna plume ozone by Surl et al. [24], for example, found 11 ppbv ozone depletion measured in plume exposure of 9 ppmv SO 2 at just 300 m downwind. Surl et al. [24] also suggest that ozone in the Mt Etna crater-rim plume was depleted below background, finding from their near-source field-campaign that ∆O 3 /∆SO 2 = (−1.02 ± 0.07) × 10 −5 s −1 with an intercept of (−6.2 ± 0.05) × 10 −4 mol/mol. A much greater depletion of tropospheric ozone (tens of ppbv ozone loss in plume of 1 ppmv SO 2 ) was observed by Kelly et al. [23] through spatial mapping of the Mt Redoubt 2009 eruption plume. The ∆O 3 /∆SO 2 data from Kelly et al. [23] is shown in Figure 5 as a function of distance downwind for two flight campaigns. A smaller ozone depletion was observed in August compared to June 2009. We focus our comparison on the June 2009 data for which the volcanic halogen emission was also quantified (just one day prior to the aircraft campaign) that yields a linear regression in ∆O 3 /∆SO 2 of (−7.2 ± 1.3) × 10 −3 mol/mol km −1 , with intercept close to zero (within statistical uncertainty). Table 2 summarizes these field measurements and interconverts ∆O 3 /∆SO 2 per unit travel time or per distance downwind using available or estimated wind-speeds. Not shown in Table 2 are aircraft studies of the 2010 Eyjafjallajökull eruption plume over Europe [36,37] that found large ozone depletions in the far-field dispersed plume but which are difficult to interpret because the halogen emission was poorly constrained and the assumption of SO 2 as a quasi-plume-tracer (in ∆O 3 /∆SO 2 ) may not be valid.
Geosciences 2018, 7, x FOR PEER REVIEW 9 of 14 the observations provide a valuable end-member example that quantifies ozone in the plume of a low halogen emitter. This contrasts to studies to date that mostly focused on volcanic plumes with higher halogen contents. To interpret the field-data, ozone depletion is scaled relative to co-measured SO2 in order to distinguish between chemistry and plume dilution effects, and it is interpreted in terms of distance or travel time downwind and the bromine emission. For the Kīlauea plumes, ozone was shown to be close to ambient levels during plume exposures of up to 10 ppmv SO2 at the Halema'uma'u crater-rim, and they were only slightly depleted in the grounding plume at around 10 km downwind from Pu'u 'Ō'ō (e.g., by up to 2 ppbv measured at 5-8 ppmv SO2). The measurements indicate that the volcano emission has only a limited impact on tropospheric ozone, as might be expected for a low halogen emitter. This contrasts with findings for higher halogen emitters. A study of Mt Etna plume ozone by Surl et al. [24], for example, found 11 ppbv ozone depletion measured in plume exposure of 9 ppmv SO2 at just 300 m downwind. Surl et al. [24] also suggest that ozone in the Mt Etna crater-rim plume was depleted below background, finding from their near-source field-campaign that ΔO3/ΔSO2 = (−1.02 ± 0.07) × 10 −5 s −1 with an intercept of (−6.2 ± 0.05) × 10 −4 mol/mol. A much greater depletion of tropospheric ozone (tens of ppbv ozone loss in plume of 1 ppmv SO2) was observed by Kelly et al. [23] through spatial mapping of the Mt Redoubt 2009 eruption plume. The ΔO3/ΔSO2 data from Kelly et al. [23] is shown in Figure 5 as a function of distance downwind for two flight campaigns. A smaller ozone depletion was observed in August compared to June 2009. We focus our comparison on the June 2009 data for which the volcanic halogen emission was also quantified (just one day prior to the aircraft campaign) that yields a linear regression in ΔO3/ΔSO2 of (−7.2 ± 1.3) × 10 −3 mol/mol km −1 , with intercept close to zero (within statistical uncertainty). Table 2 summarizes these field measurements and interconverts ΔO3/ΔSO2 per unit travel time or per distance downwind using available or estimated wind-speeds. Not shown in Table 2 are aircraft studies of the 2010 Eyjafjallajӧkull eruption plume over Europe [36,37] that found large ozone depletions in the far-field dispersed plume but which are difficult to interpret because the halogen emission was poorly constrained and the assumption of SO2 as a quasi-plumetracer (in ΔO3/ΔSO2) may not be valid.    Figure 6 presents the ∆O 3 /∆SO 2 scaled to distance or travel time downwind (i.e., rate of ozone depletion) as a function of Br/S in the volcano emission for the three case studies of Table 2 (Kīlauea, Mt Etna, and Mt Redoubt). The data show that ozone depletion is a non-linear function of the bromine emission; greater ozone loss occurs for a higher Br/S emission, as expected. However, the rate of ozone depletion for the volcano with the highest Br/S emission (Mt Redoubt) is disproportionately slow, i.e., the Mt Redoubt plume is found to be less efficient at destroying tropospheric ozone than would be expected. This non-linearity is initially a surprising result given the role of volcanic bromine in the depletion of tropospheric ozone through a "bromine explosion" that is autocatalytic (i.e., self-enhancing). A potential reason could be the complete titration of plume ozone (thereby preventing any further ozone depletion), but this is ruled out based on the measurements of Kelly et al. [23]; ozone in the Mt Redoubt plume was reduced to less than ambient levels but was not fully depleted. Instead, an explanation is provided by atmospheric box modelling of the near-source plume chemistry. Model studies point to non-linearities in the conversion of emitted HBr into reactive bromine species, depending on the bromine content of the emissions [24,29]. For high Br/S emissions, not all of the emitted HBr may become converted into reactive bromine. This was shown in a model sensitivity study by Roberts et al. [29] for Mt Etna that predicts HBr is rapidly and fully converted into reactive forms over tens of minutes for an emission with HBr/SO 2 molar ratio of 7.4 × 10 −4 , but HBr is more slowly and only partially (~50% after one hour) converted into reactive bromine for an emission with a higher HBr/SO 2 molar ratio of 2.4 × 10 −3 . Furthermore, the model simulation of the Mt Redoubt 2009 eruption plume by Kelly et al. [23] predicts that only 30% of emitted HBr was converted into reactive forms, for an emission with even higher HBr/SO 2 molar ratio of 4.1 × 10 −3 . However, another difference between the plumes is the emission flux, with 4.3 kg/s SO 2 flux for Mt Redoubt and~20 kg/s SO 2 flux for the Mt Etna simulation. The net conversion of emitted volcanic HBr into reactive bromine reflects the balance of plume chemistry processes (that form reactive halogens from HBr and that can re-form HBr from reactive halogens) as a function of plume properties such as gas flux, aerosol, plume-air mixing, etc., and is also a function of time. For plumes younger than one minute, recent observations by Rüdiger et al. [45] suggest reactive bromine accounts for less than 44% of total bromine, consistent with the models. The modeled plume Br-speciation predicted by Roberts et al. [23] and [29] are shown in Figure 7, illustrating the proportion of emitted HBr converted into reactive bromine during the first hours of plume chemical evolution. These model outputs are used to re-evaluate the non-linear trend in Figure 6. When the Br/S content of the emissions is adjusted to reflect the modeled plume "reactive bromine"/S (e.g., reduced to 30% of the total emitted Br/S for Mt Redoubt: open triangle, Figure 6), there is more linearity (broadly proportional trend) between the volcanic (reactive) bromine/S and the rate of plume ozone depletion. It is nevertheless expected that other variables, e.g., aerosol and plume-air mixing rate, may affect ∆O 3 /∆SO 2 depending on the volcanic-meteorological setting.
It is nevertheless expected that other variables, e.g., aerosol and plume-air mixing rate, may affect ΔO3/ΔSO2 depending on the volcanic-meteorological setting. Figure 6. Depletion in ozone scaled to volcanic SO2 (as a quasi-plume-tracer) and scaled with (i) km distance downwind or (ii) seconds travel time downwind, plotted as a function of the bromine/SO2 molar ratio in the emission. Data are available for three volcanic systems: Kīlauea (red circle, this study), Mt Etna (purple square, [24]), and Mt Redoubt (black triangle, [23]), in ascending order of Br/S in their emissions. Arrow and open triangle denote adjustments in the Br/S content that accounts for partial rather than full conversion of emitted HBr into reactive bromine. See text for details.  [29], and Kelly et al. [23] illustrating varying degrees (complete, partial) of conversion of HBr into reactive bromine. Reproduced with permission from Kelly et al. [23], JVGR; published by Elsevier. Figure 6. Depletion in ozone scaled to volcanic SO 2 (as a quasi-plume-tracer) and scaled with (i) km distance downwind or (ii) seconds travel time downwind, plotted as a function of the bromine/SO 2 molar ratio in the emission. Data are available for three volcanic systems: Kīlauea (red circle, this study), Mt Etna (purple square, [24]), and Mt Redoubt (black triangle, [23]), in ascending order of Br/S in their emissions. Arrow and open triangle denote adjustments in the Br/S content that accounts for partial rather than full conversion of emitted HBr into reactive bromine. See text for details. It is nevertheless expected that other variables, e.g., aerosol and plume-air mixing rate, may affect ΔO3/ΔSO2 depending on the volcanic-meteorological setting.    [29], and Kelly et al. [23] illustrating varying degrees (complete, partial) of conversion of HBr into reactive bromine. Reproduced with permission from Kelly et al. [23], JVGR; published by Elsevier.
In conclusion, chemistry in the near-source volcanic plume exerts an important influence on the fate of volcanic halogens entering the troposphere. Plume chemistry converts the volcanic HBr emission into reactive bromine, causing the depletion of tropospheric ozone. Measurements of ozone in the near-source volcanic plume are challenging to make but they can provide useful observational constraints on the volcanic plume halogen chemistry. This study presents observations of ozone in the volcanic plumes from Kīlauea, a low-halogen emitter. Ozone was close to ambient levels in crater-rim Halema'uma'u emissions and only slightly depleted in the plume 10 km downwind from Pu'u 'Ō'ō. This contrasts with observations of much larger ozone depletion (tens of ppbv) in the near-source and near-downwind plumes of two higher halogen emitters: Mt Etna and Mt Redoubt [23,24]. The available observations combined with numerical modelling of the plume chemistry suggest that the volcanic Br/S emission and its (partial or complete) conversion into reactive bromine are key controls in the depletion of tropospheric ozone downwind from the volcano. Characterizing the near-source volcanic plume chemistry is thus an essential step to quantifying the fate and downwind impacts of volcanic emissions to the atmosphere.